I code and sell minimilistic ai video software
- Thread starter Pythy
- Start date
It can be downloaded here:
I hope to get a feedback on what you think.
Smilingreen
Known around here
"TURN ORDINARY CAMERAS
TO A RING,NEST,ARLO,WYZE USING
AN ARTIFICIAL INTELLIGENCE SOFTWARE"
Just to ask, why would I want to down-grade my cams?
TO A RING,NEST,ARLO,WYZE USING
AN ARTIFICIAL INTELLIGENCE SOFTWARE"
Just to ask, why would I want to down-grade my cams?
Smilingreen
Known around here
So, in looking at one of your YT vids, your mask detection software appears to have a minimum of a 40% false negative failure. Out of 5 people in the video that the software detected, 2 of them were compliant and had masks on, but the software shows them as not compliant. Care to explain how this would be acceptable?
Well I do admit no model is actually perfect . But the false negatives are extremely low. I hope you watched all the video. Also the models were trained on 1080p pixel like images. This video in particular had poor quality. But overall the model is robust and can be retrained with more data to be better. Thank you.
I get your point. This particular software main point is it's ai people detection. It had been trained to detect people at high accuracy thus reducing false positives.The main idea was to connect it to an existing camera which gives alot of false positives.
I will be happy if you can give me a feedback on what you think about the software.
Thank you
Smilingreen
Known around here
I did watch them and I did notice in several other clips, the AI at times, was having issues distinguishing between masks and no masks. As far as the software, I run OSX, so I don't believe you have a version that will run on Macs. One of the PC guys should be able to help you with that.
[/QUOTE]
Yeah I recognize that part of the video. It was a bit tricky. The mask was not worn well. Ai was trained mostly on datasets containing images of with mask or without mask.So it was oscillating between he has o he has not. However with alot of datasets which contains eg mask_not_worn_well' , the ai would have picked it up.
The code can actually be compiled for Mac ,but unfortunately I do not have a Mac .
Thank you
Yeah I recognize that part of the video. It was a bit tricky. The mask was not worn well. Ai was trained mostly on datasets containing images of with mask or without mask.So it was oscillating between he has o he has not. However with alot of datasets which contains eg mask_not_worn_well' , the ai would have picked it up.
The code can actually be compiled for Mac ,but unfortunately I do not have a Mac .
Thank you